Stay Tuned: An Empirical Study of the Impact of Hyperparameters on LLM Tuning in Real-World Applications.
Alon HalfonShai GretzOfir ArvivArtem SpectorOrith Toledo-RonenYoav KatzLiat Ein-DorMichal Shmueli-ScheuerNoam SlonimPublished in: CoRR (2024)
Keyphrases
- hyperparameters
- model selection
- cross validation
- bayesian framework
- random sampling
- closed form
- bayesian inference
- prior information
- gaussian process
- support vector
- parameter settings
- maximum a posteriori
- incremental learning
- maximum likelihood
- em algorithm
- sample size
- gaussian processes
- noise level
- regularization parameter
- machine learning
- incomplete data
- training set
- data mining
- missing values
- computer vision
- parameter optimization